Original Article
Active contouring and 3D model deformable registration of radiotherapy planning and cone-beam computed tomography images
Abstract
Background: To ensure high accuracy during radiation therapy (RT), the image-guided RT (IGRT) technique uses on-board cone-beam computed tomography (CBCT) scanning as an image guidance procedure for target localization before and during treatment. Adaptive RT aiming to modify RT target volumes according to kinetic changes in tumor shape during RT course is based on registration of CBCT and planning CT images. However, the re-contouring and re-planning procedures are extensively time and cost consuming. We developed a novel automatic contouring and image registration method to replace the manual re-contouring with accurate image registration.
Methods: For the image sets with format of Digital Imaging and Communications in Medicine (DICOM) standard, we wrote a program in MATLAB language (Version R2016a) to read and convert CBCT images into cross-sectional (tomographic) images similar to those obtained via planning CT. For image enhancement, the active contouring by using Chan-Vese model with level set formulation was applied. To overcome the variations in spatial location of these two sets of CT images, the iterative closest point (ICP) algorithm was used for 3D model registration. The deformable image registration (DIR) with Double force Demons algorithm was performed for auto-transformation of contours from planning CT to CBCT images.
Results: The customized program accurately converted the format of CBCT to planning CT. Image enhancement was achieved by our modified active contour model which solved the energy minimization problem. In 3D model registration, the variations in spatial location of the CBCT and planning CT images were corrected. After selection of most similar images, the planning CT images were registered to corresponding CBCT images. The registered images were clearer than CBCT images with removal of other confounding structures outside body contours.
Conclusions: The planning CT and CBCT images could be precisely registered by using a novel established technique consisting of active contouring with 3D model and DIR. This technique would enable the on-line radiation treatment planning for adaptive radiotherapy.
Methods: For the image sets with format of Digital Imaging and Communications in Medicine (DICOM) standard, we wrote a program in MATLAB language (Version R2016a) to read and convert CBCT images into cross-sectional (tomographic) images similar to those obtained via planning CT. For image enhancement, the active contouring by using Chan-Vese model with level set formulation was applied. To overcome the variations in spatial location of these two sets of CT images, the iterative closest point (ICP) algorithm was used for 3D model registration. The deformable image registration (DIR) with Double force Demons algorithm was performed for auto-transformation of contours from planning CT to CBCT images.
Results: The customized program accurately converted the format of CBCT to planning CT. Image enhancement was achieved by our modified active contour model which solved the energy minimization problem. In 3D model registration, the variations in spatial location of the CBCT and planning CT images were corrected. After selection of most similar images, the planning CT images were registered to corresponding CBCT images. The registered images were clearer than CBCT images with removal of other confounding structures outside body contours.
Conclusions: The planning CT and CBCT images could be precisely registered by using a novel established technique consisting of active contouring with 3D model and DIR. This technique would enable the on-line radiation treatment planning for adaptive radiotherapy.
